Leveraging Instance-, Image- and Dataset-Level Information for Weakly Supervised Instance Segmentation

Open paper
Benchmark Model Rank Results
image-level-supervised-instance-segmentationLIID#6mAP@0.5: 48.4mAP@0.75: 24.9
image-level-supervised-instance-segmentation-1LIID#4AP: 16.0AP@50: 27.1AP@75: 16.5
weakly-supervised-instance-segmentation-onLIID#2Average Best Overlap: 50.8mAP@0.5: 48.4mAP@0.75: 24.9
weakly-supervised-semantic-segmentation-onLIID (Res2Net-101)#49Mean IoU: 69.4
weakly-supervised-semantic-segmentation-onLIID (ResNet-101, +24K SI)#60Mean IoU: 67.8
weakly-supervised-semantic-segmentation-onLIID (ResNet-101)#66Mean IoU: 66.5
weakly-supervised-semantic-segmentation-on-1LIID#57Mean IoU: 67.5

Rank counts only results with a code link.